Professional Documents
Culture Documents
Sentinel-2A
Geometric benchmarking over
Maussane test site for CAP
purposes
Blanka Vajsova
Pär Johan Åstrand
2015
JRC99517
EUR 27674 EN
ISBN 978-92-79-54237-4
ISSN 1831-9424
doi:10.2788/544302
All images © European Union 2015, except: page 5, European Space Agency, image 1], Year. Source: [ref. viii]
The geographic borders are purely a graphical representation and are only intended to be indicative. The
boundaries do not necessarily reflect the official position of the European Commission.
How to cite: Vajsová B, Aastrand P.; New sensors benchmark report on Sentinel-2A; EUR 27674 EN;
doi:10.2788/544302
Table of contents
Abstract ................................................................................................................ 3
1 Introduction ................................................................................................ 4
1.1 Objective ................................................................................................. 4
2 Sentinel-2 mission ....................................................................................... 4
2.1 Satellite sensor characteristics – design ....................................................... 5
2.2 Satellite sensor characteristics – specifications ............................................. 5
2.3 Sentinel-2 image products available to users ................................................ 6
2.4 Sentinel-2 Geometric Quality Requirements ................................................. 6
2.5 Sentinel-2 Data format .............................................................................. 6
3 Maussanne – test site ................................................................................... 7
4 Input data ................................................................................................... 7
4.1 Independent check points (ICPs) ................................................................ 7
5 Sentinel-2 testing dataset ............................................................................. 8
5.1 1C level image product .............................................................................. 8
5.2 Global Reference Image ............................................................................. 9
6 External quality control ................................................................................. 9
6.1 External quality control methodology ........................................................... 9
6.2 Outcome ................................................................................................ 11
6.2.1 Absolute geometric accuracy ..................................................................... 11
6.2.2 Relative geometric accuracy ...................................................................... 12
6.3 Discussion .............................................................................................. 14
7 Conclusions and prospects .......................................................................... 14
References .......................................................................................................... 16
List of abbreviations and definitions ........................................................................ 18
List of figures: ..................................................................................................... 19
List of tables ........................................................................................................ 20
List of Annexes .................................................................................................... 20
2
Abstract
The main objective of the report is to assess whether the Sentinel-2 sensor can be
qualified for Control with Remote Sensing programme, specifically in the Common
Agriculture Policy (CAP) Controls image acquisition campaign. The benchmarking
presented herein aims at evaluating the usability of Sentinel-2 for the CAP checks
through an estimation of its geometric (positional) accuracy.
For that purpose, the External Quality Control of Sentinel-2 orthoimagery conforms to
the standard method developed by JRC and follows a procedure already adopted in the
validation of previous high and very-high resolution products.
3
1 Introduction
The Common Agriculture Policy (CAP) uses the “Controls with Remote Sensing” (CwRS)
as one of control systems to check whether aids given to European farmers are correctly
granted. The JRC service together with chosen Image Providers (IPs) assures both, a
smooth acquisition of appropriate image data and their initial quality assessment.
Each newly launched satellite with an ambition to provide image data for the purpose of
CAP checks has to pass a validation test to prove a fulfilment of CwRS requirements [ref.
ii, iii].This geometric validation is based on the External Quality Control (EQC) of the
orthoimagery and follows strict guidelines described by JRC in the so-called "Guidelines
for Best Practice and Quality Checking of Ortho Imagery" [ref. i].
Within this context, the purpose of the current study is to perform an initial quality
assessment with respect to the capabilities of the newly launched Sentinel-2A satellite,
see chapter 2.
Namely, the sensor requirement implies that the planimetric accuracy of the
orthoimagery, expressed as the Root-Mean-Square Error (RMSE) in Easting and Northing
directions, should not exceed 15m to fulfill the geometric requirements and
specifications of HR prime profile and HHR ortho profile defined in the HR profile based
technical specifications for the CAP checks.
1.1 Objective
The aim of this report is to summarize the outcome of the geometric quality testing of
the Sentinel-2 images acquired over the Joint Research Centre (JRC) Maussane test site.
The objective of this study is twofold:
to evaluate the planimetric accuracy of the orthorectified Sentinel-2 imagery;
to check if the orthorectified imagery of the Sentinel-2A meet the CAP CwRS
Programme technical requirements (see Chapter 7) [ref. ii, iii].
2 Sentinel-2 mission
Sentinel-2 is an Earth observation mission developed by European Space Agency (ESA)
as part of the Copernicus Programme to perform terrestrial observations in support of
services such as forest monitoring, land cover changes detection, and natural disaster
management, humanitarian relief operations, risk mapping and security concerns. It
consists of two identical satellites, Sentinel-2A and Sentinel-2B providing continuity for
the current SPOT and Landsat missions. The two satellites will work on opposite sides of
the orbit. The launch of the first satellite, Sentinel-2A, occurred 23 June 2015 on a Vega
launch vehicle. Sentinel-2B will be launched in mid-2016 [ref. iv].
The mission provides a global coverage of the Earth's land surface every 10 days with
one satellite and 5 days with 2 satellites, making the data of great use in on-going
studies.
The satellites are equipped with the state-of-the-art Multispectral Imager (MSI)
instrument that offers high-resolution optical imagery. This MSI imager uses a push-
broom concept and its design has been driven by the large 290 km swath requirements,
together with the high geometrical and spectral performance required of the
measurements [ref. v]
As a prime contractor to construct the Sentinel-2 satellite has been appointed Astrium
Germany, leading also a consortium with core partners [ref. v]:
Astrium France is providing the MSI payload
Boostec is providing the three-mirror Silicon carbide telescope and the instrument
baseplate
Jena-Optronik is responsible for the 2 Video Compression Units (VCU)
Sener is supplying the instrument Calibration and Shutter Mechanism (CSM).
4
2.1 Satellite sensor characteristics – design
Date: June 23, 2015
Launch Vehicle: Vega rocket
Launch information
Launch Location: Europe’s Spaceport near Kourou in
French Guiana
Satellite weight/size/power approx. 1200 kg; 3.4 m x 1.8 m x 2.35 m; 1.7kW
Altitude: 786 km
Orbit Type: sun-synchronous
Period: min
Inclination/Equator Crossing 98.62 deg/ 10:30pm (ascending node)
Time
Orbits per day 14.3 revolutions per day
Revisit rate 10 days with one satellite and 5 days with 2 satellites
Operational lifespan 7.25 years (with consumables for 12)
all continental land surfaces (including inland waters)
between latitudes 56° south and 83° north
all coastal waters up to 20 km from the shore
Coverage
all islands greater than 100 km2
all EU islands, the Mediterranean Sea
all closed seas (e.g. Caspian Sea)
Table 1: Sentinel-2 mission - design
5
2.3 Sentinel-2 image products available to users
6
3 Maussanne – test site
The geometric quality assessment of the Sentinel-2A image data has been performed
over a standard test site of Maussane, located in French commune Maussane-les-Alpilles
in the Provence-Alpes-Cote d’Azur region in southern France, see Figure 2.
The site contains a low mountain massif, mostly covered by forest, surrounded by low
lying agricultural plains and a lot of olive groves. A number of low density small urban
settlements and a few limited water bodies are present over the site [ref. xi].
The site has been used by JRC for the geometric benchmarking of High Resolution (HR)
and Very High Resolution (VHR) imagery since 1997 for the following reasons [ref. x]:
it presents a variety of agricultural conditions typical for the EU, as well as urban
settlements and water bodies,
it contains a low mountain massif (650m above sea level) mainly covered by
forest, surrounded by agricultural areas.
during the years, a time series of reference data (i.e. DEMs, imagery, ground
control points) has been collected. Altogether there are available 8 GCP datasets
(292 points) of various positional accuracies, see Table 4 and Figure 3.
4 Input data
7
As regards to the positional accuracy of ICPs, according to the Guidelines (Kapnias et al.,
2008) [ref. i] the ICPs should be at least 3 times more precise than the target
specification of the orthoproduct, i.e. in our case of a target 15 m RMS error the ICPs
should have a specification of 5.0m (3m recommended). All ICPs that have been selected
fulfil therefore the defined criteria (Table 4).
Figure 3: Maussane test site and related available JRC ancillary data: DEM and CPs.
The ADS40 DEM covering a large extent (35 x20 km²) over Maussane area is diplayed as a background
grayscale layer (the brighter a pixel, the higher the elevation at that point). Over that same area, 8 datasets of
CPs are retrieved from previous campaigns [ref. x] and represented as coloured dots on the figure. For
geometric specifications of each dataset see the Table 4. The footprints of the two test areas are represented
as coloured frames.
For the evaluation of the geometric accuracy of the Sentinel-2 ortho imagery, 15 to 21
independent ICPs were selected by a JRC operator. Considering the accuracy,
distribution and recognisability on the given images, points from the seven datasets
were decided to be used for the EQC, see Table 4.
8
5.2 Global Reference Image
In order to meet the multi-temporal registration and the absolute geolocation
requirements, a Global Reference Image (GRI) will be generated and used for the
automatic extraction of GCPs for the systematic refinement of the geometric model at
the end of the Level-1B processing.The database will be a composite of cloud-free (or
with a limited presence of clouds), geometrically refined and mono-spectral (the current
baseline is to use Band 4) Level-1B granules/datastrips covering a full repeat cycle (143
orbits, i.e. 10 days of acquisition).The GRI will be gradually completed (as the images
become available all around the world) through an appropriate selection of Level-1B
images followed by an accurate geometric refinement performed on the basis of spatio-
triangulation [ref. ix]
The spatio-triangulation process is based on a bundle adjustment of set of images
combined with orbit information and GCPs refinement.
where X,YREG(i) are ortho imagery derived coordinates, X,Y(i) are the ground true
coordinates, n express the overall number of ICPs used for the validation.
This geometric accuracy representation is called the positional accuracy, also referred to
as planimetric/horizontal accuracy and it is therefore based on measuring the residuals
between coordinates detected on the orthoimage and the ones measured in the field or
on a map of an appropriate accuracy [ref. xiii].
9
Figure 4: JRC ancillary data – visibility of ICPs on Sentinel-2 images
8 datasets of ICPs retrieved from previous campaigns [ref.x] and represented as coloured dots on the figure.
Red colour represents a point not well identifiable, an orange coloure a medium identifiability and a green
colour means that the point is on the image well visible and identifiable. For geometric specifications of each
dataset see Table 4. The footprints of the two test areas are represented as coloured frames.
To provide accurate and reliable results two separate test AOIs were selected:
“The small AOI”, covering an extent of 10x10 km2, with UL corner at position
(636225E, 4846850N) in EPSG 32631 (UTM zone 31N, ellipsoid WGS84)
reference system, this AOI is usually used for VHR sensors benchmarking,
therefore corresponding auxiliary image data are available ( WV2, WV3. GE1,
Pleiades..). See the blue box in the Figure 4.
“The big AOI” covering an extent of 19x18 km2 (East x North) with UL corner at
position (648800E, 4854500N) in EPSG 32631 (UTM zone 31N, ellipsoid WGS84)
reference system, usually used for HR sensors benchmarking with corresponding
auxiliary image data (SPOT5,6,7, RE, THEOS..). See the green box in the Figure
4.
To support the absolute geometric accuracy results calculated on the basis of ground
true coordinates (measured in the field), also the relative geometric accuracy was
considered.
The following ortho products were used as reference data:
WV3 ortho image of max RMSE of 0.60m and pixel size of 0.40m, covering “the
small AOI”
SPOT 7 ortho image of max RMSE of 4.50m and pixel size of 0.1.5m, covering
“the big AOI”
10
Concerning the relative geometric accuracy two different approaches for the ICPs
selection were applied. The classic manual ICPs collection was complemented with an
automatic correlation of ICPs.
For the automatic ICPs generation the Image AutoSync module of ERDAS IMAGINE was
used, particularly the automatic point matching (APM) function.
The APM is a software tool that uses image-matching technology to automatically
recognize and measure the corresponding image points between two raster images. In
IMAGINE AutoSync, APM aims to deliver the coordinates of evenly distributed
corresponding points between an input image (Sentinel-2A) and a reference image
(SPOT 7, WV3) [ref. xvi]
The APM tool matches the control points by making use of a pyramid data structure to
match level by level. When APM begins to run, firstly, it establishes respectively a 3×3
image pyramid data structure for the input image and the reference image, which is a
group of image sequences generated from the low to high resolution. It begins to match
from the lowest level of resolution. The APM finds the matching point and maps it to the
search area of the last layer. Then it improves the resolution layer of both images and
matches again in the search area. The cycle repeats until reaching the original image
resolution. The matching points of the two images are obtained [xiv]. Further
information about accuracy analysis of this module can be found in [ref. xv].
6.2 Outcome
6.2.1 Absolute geometric accuracy
Figure 5: ICPs dataset used by JRC in the EQC of Sentinel-2A ortho imagery
The blue frame on the left represents AOI 10mx10m where 15 ICPs were selected. The green frame on the
right represents AOI 19mx18km where 21 ICPs were used for testing.
11
6.2.2 Relative geometric accuracy
Figure 6: ICPs selected by JRC for the EQC (relative accuracy) of Sentinel-2A ortho
imagery
In both areas 21 ICPs were selected.
Since the absolute positions (e.g. DGPS measurement) of these check points are not
known, the validation results can be interpreted as relative values to the reference ortho
images, i.e. WV3 or SPOT 7 ortho image accuracy. The geometric characteristics of the
WV3 image, and in particular its spatial resolution, are significantly better that the being
studied Sentinel-2, therefore (only within this context) the ICPs coordinates measured
on WV3 ortho image could be even treated as the absolute coordinates.
12
6.2.2.2 Automatic correlation of ICPs
Matching the Sentinel 2A data with the WV3 reference image give less satisfactory
results (higher values of RMSEs, much less ICP) due to the huge difference between the
resolutions of the sensors.
The resolution creates a difference in the details of the two images. It is recommended
to avoid mixing input and reference images with a resolution difference larger than a
factor of six [ref xvi]. WV3 as a reference image apparently does not adhere to the
suggestion (the resolution of the WV3 is 25x better than the Sentinel-2A one). There are
substantial differences between the RMSEs of the images. To follow the recommendation
we decided not to include the results into the summary.
13
6.3 Discussion
9.0
8.0 Relative RMSE related to SPOT 7
7.0 (manual)
RMSE X [m]
All calculated RMSEs resulted below one pixel. Regarding the absolute accuracy, the
RMSEs of both tested AOIs were comparable and thus the final RMSE was calculated as
an average.
The relative geometric accuracy values supported good absolute geometric accuracy
results.
The automatic ICPs correlation is not limited to human visual interpretation and it is not
so work intensive as manual point measurement. The output of the automatic point
matching algorithm is better in accuracy in comparison to the current methodology,
however the attention has to be paid to a suitability of a reference image (resolution,
selected band, time of capture…) and APM strategy parameters.
In the medium-term, geometric refinement using the Global Reference Image should
further increase the geometric quality of the Sentinel-2A products.
The Sentinel-2A data are available to all users via the Scientific data Hub:
https://scihub.esa.int/
14
ANNEX A
Image id S2A_OPER_MTD_L1C_TL_MTI__20
(internal image id) 150819T203140_A000820_T31TFJ
Image short ID S2A_820
Product level Level 1C
Product Type MSP
Collection date 19/8/2015
Ellipsoid Type/Projection WGS-84/UTM, N31
Format JPEG 2000
Bits Per Pixel 12
Image id S2A_OPER_MTD_L1C_TL_MTI__20
(internal image id) 150822T204401_A000863_T31TFJ
Image short ID S2A_863
Product level Level 1C
Product Type MSP
Collection date 22/8/2015
Ellipsoid Type/Projection WGS-84/UTM, N31
Format JPEG 2000
Bits Per Pixel 12
Image id S2A_OPER_MTD_L1C_TL_MTI__20
(internal image id) 150829T203120_A000963_T31TFJ
Image short ID S2A_963
Product level Level 1C
Product Type MSP
Collection date 29/8/2015
Ellipsoid Type/Projection WGS-84/UTM, N31
Format JPEG 2000
Bits Per Pixel 12
Image id S2A_OPER_MTD_L1C_TL_MTI__20
(internal image id) 150908T203133_A001106_T31TFJ
Image short ID S2A_1106
Product level Level 1C
Product Type MSP
Collection date 8/9/2015
Ellipsoid Type/Projection WGS-84/UTM, N31
Format JPEG 2000
Bits Per Pixel 12
Image id S2A_OPER_MTD_L1C_TL_MTI__20
(internal image id) 150918T204543_A001249_T31TFJ
Image short ID S2A_1249
Product level Level 1C
Product Type MSP
Collection date 18/9/2015
Ellipsoid Type/Projection WGS-84/UTM, N31
Format JPEG 2000
Bits Per Pixel 12
15
References
i. Kapnias, D., Milenov, P., Kay, S. (2008) Guidelines for Best Practice and Quality
Checking of Ortho Imagery. Issue 3.0. Ispra
ii. JRC IES, VHR image acquisition specifications for the CAP checks (CwRS and LPIS
QA), VHR profile-based specifications including VHR+ profiles (2015, 2016),
available at
https://g4cap.jrc.ec.europa.eu/g4cap/Portals/0/Documents/17359.pdf
https://g4cap.jrc.ec.europa.eu/g4cap/Portals/0/Documents/21449_21112015_final
.pdf
iii. JRC IES, HR image acquisition specifications for the CAP checks (CwRS), HR profile
-based specifications (2015, 2016), available at
https://g4cap.jrc.ec.europa.eu/g4cap/Portals/0/Documents/17362.pdf
https://g4cap.jrc.ec.europa.eu/g4cap/Portals/0/Documents/21450_21112015_final
.pdf
iv. https://earth.esa.int/web/sentinel/missions/sentinel-2
v. https://earth.esa.int/web/guest/missions/esa-operational-eo-missions/sentinel-2
vi. https://earth.esa.int/web/sentinel/user-guides/sentinel-2-msi
vii. Sentinel-2 User Handbook, ESA Standard Document, 24/07/2015 Issue 1 Rev 2
viii. François Spoto , Philippe Martimort, Omar Sy and Paolo Laberint, Sentinel-2 Project
team, ESA/ESTEC, Sentinel-2 Optical High Resolution Mission for GMES Operational
services, Sentinel-2 preparatory symposium, 23-27 April 2012, ESA-ESRIN,
Frascati(Rome) Italy, available at
http://www.congrexprojects.com/docs/12c04_doc/4sentinel2_symposium_spoto.pd
f
ix. Sentinel-2 PDGS Project Team, Sentinel-2 Calibration and Validation Plan for the
Operational Phase, 22 December 2014
x. Nowak Da Costa, J., Tokarczyk P., 2010. Maussane Test Site Auxiliary Data:
Existing Datasets of the Ground Control Points.
xi. Lucau, C., Nowak Da Costa J.K. (2009) Maussane GPS field campaign: Methodology
and Results.Available at
http://publications.jrc.ec.europa.eu/repository/bitstream/111111111/14588/1/pub
sy_jrc56280_fmp11259_sci-tech_report_cl_jn_mauss-10-2009.pdf
xii. Grazzini, J., Astrand, P., (2013). External quality control of SPOT6. Geometric
benchmarking over Maussane test site for positional accuracy assessment
orthoimagery. Available at
http://publications.jrc.ec.europa.eu/repository/bitstream/111111111/29232/1/lb-
na-26-103-en-n.pdf
16
Remote Sensing Vol. 2 No. 2 (2013) , Article ID: 33180 , 9
pages,DOI:10.4236/ars.2013.22018
xvi. ERDAS, Inc., IMAGINE AutoSync™ User’s Guide September 2008
17
List of abbreviations and definitions
AD Attitude Determination
ADS Airborne Digital Sensor
AOI Area of Interest
APM Automatic Point Matching
CAP The Common Agricultural Policy
CE90 Circular Error of 90%
COTS Commercial off-the-shelf
CSM Calibration and Shutter Mechanism
DEM Digital Elevation Model
DSM Digital Surface Model
EO Earth Observation
EPSG European Petroleum Survey Group
EQC External Quality Control
ESA European Space Agency
GCP Ground Control Point
GRI Global Reference Image
GPS The Global Positioning System
GSD Ground Sample Distance
HR High resolution
IPC Independent Check Point
IQC Internal Quality Control
JRC Joint Research Centre
LE90 Linear Error of 90%
LPIS Land Parcel Information System
LVLH Local Vertical/Local Horizontal
MS Multispectral
MSI Multispectral Imager
OD Orbit Determination
ONA Off Nadir Angle
PAN Panchromatic
POD Precision Orbit Determination
RMSE Root Mean Square Error
RPC Rational Polynomial Coefficient
SAR Synthetic-Aperture Radar
TP Tie Point
UTM Universal Transverse Mercator
VCU Video Compression Units
VHR Very High Resolution
WGS 84 World Geodetic System 1984
1-D One-dimensional
18
List of figures:
19
List of tables
List of Annexes
20
Europe Direct is a service to help you find answers to your questions about the European Union
Free phone number (*): 00 800 6 7 8 9 10 11
(*) Certain mobile telephone operators do not allow access to 00 800 numbers or these calls may be billed.
A great deal of additional information on the European Union is available on the Internet.
It can be accessed through the Europa server http://europa.eu
21
LB-NA-27674-EN-N
JRC Mission
As the Commission’s
in-house science service,
the Joint Research Centre’s
mission is to provide EU
policies with independent,
evidence-based scientific
and technical support
throughout the whole
policy cycle.
Working in close
cooperation with policy
Directorates-General,
the JRC addresses key
societal challenges while
stimulating innovation
through developing
new methods, tools
and standards, and sharing
its know-how with
the Member States,
the scientific community
and international partners.
Serving society
Stimulating innovation
Supporting legislation
doi:10.2788/544302
ISBN 978-92-79-54237-4
22